Some performance figures taken from the Nexus S. These are from An3DBenchXL. The tests are all running fine, just the multi-texturing in the Ninja test isn't supported right now, so it will get worse...
Using the OpenGL ES 1.1 path:
05-25 23:29:24.995: INFO/jPCT-AE(6997): Double dragon: 30.55 fps
05-25 23:29:24.995: INFO/jPCT-AE(6997): Flower power: 22.86 fps
05-25 23:29:24.995: INFO/jPCT-AE(6997): Ninjas' garden: 16.14 fps
05-25 23:29:24.995: INFO/jPCT-AE(6997): Emperor's new clothes: 43.58 fps
05-25 23:29:24.995: INFO/jPCT-AE(6997): Magic island: 37.33 fps
05-25 23:29:24.995: INFO/jPCT-AE(6997): TOTAL SCORE: 26053
Using the OpenGL ES 2.0 path:
05-25 23:25:07.137: INFO/jPCT-AE(6940): Double dragon: 9.25 fps
05-25 23:25:07.137: INFO/jPCT-AE(6940): Flower power: 21.52 fps
05-25 23:25:07.137: INFO/jPCT-AE(6940): Ninjas' garden: 13.61 fps
05-25 23:25:07.137: INFO/jPCT-AE(6940): Emperor's new clothes: 42.26 fps
05-25 23:25:07.137: INFO/jPCT-AE(6940): Magic island: 19.59 fps
05-25 23:25:07.137: INFO/jPCT-AE(6940): TOTAL SCORE: 18394
Why is that? Well, as you can see, high polygon tests (dragon and island) suffer the most. I guess that's because a high polygon count means a high vertex count, which means a lot of calls to the vertex shader. The most expensive part of the vertex shader is the lighting, because it requires some dot-products, which seem to be slow. If you remove the light source from the tests, it runs much faster...
So..well...this is a bit disappointing, but i actually expected something like this. Too bad, that you can't mix shaders and fixed function as you can with desktop OpenGL.